Accepted Manuscript

Homogenization of degenerate coupled fluid flows and heat transport through

porous media

Michal Bene, Igor Paanin

PII:
DOI:
Reference:

S0022-247X(16)30450-4
http://dx.doi.org/10.1016/j.jmaa.2016.08.041
YJMAA 20677

To appear in:

Journal of Mathematical Analysis and Applications

Received date:

8 May 2016

Please cite this article in press as: M. Bene, I. Paanin, Homogenization of degenerate coupled fluid flows and heat transport
through porous media, J. Math. Anal. Appl. (2016), http://dx.doi.org/10.1016/j.jmaa.2016.08.041

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are
providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting
proof before it is published in its final form. Please note that during the production process errors may be discovered which could
affect the content, and all legal disclaimers that apply to the journal pertain.

Homogenization of degenerate coupled uid ows and

heat transport through porous media
Michal Benesa,, Igor Pazaninb
a

Department of Mathematics,
Faculty of Civil Engineering,
Czech Technical University in Prague,
Th
akurova 7, 166 29 Prague 6, Czech Republic
b

Department of Mathematics,
Faculty of Science,
University of Zagreb,
Bijenicka 30, 10000 Zagreb, Croatia

Abstract
We establish a homogenization result for a fully nonlinear degenerate parabolic
system with critical growth arising from the heat and moisture ow through
a partially saturated porous media. Existence of a global weak solution of the
mesoscale problem is proven by means of a semidiscretization in time, a priori
estimates and passing to the limit from discrete approximations. After that,
porous material exhibiting periodic spatial oscillations is considered and the
two-scale convergence (as the oscillation period vanishes) to a corresponding
homogenized problem is rigorously proven.
Keywords: Nonlinear degenerate parabolic system, Homogenization,
Asymptotic analysis, Two-scale convergence, Coupled transport processes
in porous media
2000 MSC: 35K45, 35B27, 35K65

Corresponding address: Department of Mathematics, Faculty of Civil Engineering,

Czech Technical University in Prague, Th
akurova 7, 166 29 Prague 6, Czech Republic
Email address: michal.benes@cvut.cz (Michal Benes)

Preprint submitted to Elsevier

August 23, 2016

1. Introduction
This paper deals with the two-scale homogenization of a class of doubly nonlinear degenerate parabolic problems arising from coupled transport
processes in porous media.
Model problem. Let be a bounded domain in R2 , C 0,1 and let D
and N be open disjoint subsets of (not necessarily connected) such that
D = and the 1-dimensional measure of \(D N ) equals zero. Let
T (0, ) be xed throughout the paper. Let us abbreviate T = (0, T ),
DT = D (0, T ) and N T = N (0, T ) and let Y = (0, 1)2 be a periodicity
cell. Vectors and vector functions are denoted by boldface letters. Let > 0
be a small scalar parameter. From the geometrical point of view, is the
characteristic length representing the small scale variability of the porous
media. We study the homogenization of the doubly nonlinear degenerate
system indexed by the scale parameter , namely,
t b (x/, u ) + q = 0
t [b (x/, u ) + (x/) ]
= [2 (x/)( , u ) q ]
u = 0
= 0
q n = 0
n = 0
u (0) = u0
(0) = 0

in T ,

(1.1)

in T ,
in DT ,
in DT ,
in N T ,
in N T ,
in ,
in .

(1.2)
(1.3)
(1.4)
(1.5)
(1.6)
(1.7)
(1.8)

Here
q = [1 (x/)a( )u + g(x/, , u )] .

(1.9)

From the physical point of view, equations (1.1) and (1.2), respectively, represent the mass balance of moisture and the heat equation for the porous
system after the so-called Kirchho transformation. Functions u : T R
and : T R are the unknowns. n is the outer unit normal vector to
, a : R R, b : R2 R R, : R R R, g : R2 R R R2 ,
1 , 2 : R2 R, : R2 R and u0 : R and 0 : R are given
functions. Roughly speaking, the coecient functions 1 , 2 and are Yperiodic functions on R2 and similarly, b and g are Y-periodic with respect
2

to the rst variable (see the next section for precise assumptions). As
gets smaller, the coecient functions 1 , 2 , , b and g in (1.1)(1.2) oscillate more rapidly. We solve the classical homogenization problem, namely
investigate the behavior of the solution [u , ] in the limit (as 0).
Remark 1.1. The analysis presented here can be straightforwardly extended
to a setting with nonhomogeneous boundary conditions (see [10] for details).
Here we work with homogeneous boundary conditions just to simplify and
shorten the presentation and avoid unnecessary technicalities in the existence
result.
System (1.1)(1.2) can be seen as the special case of more general problem, the so-called doubly nonlinear problem, t B(x, u)A(x, u, u) = 0,
with nonlinearities in both, parabolic as well as elliptic parts. From theoretical point of view, as far as we know, no general existence, uniqueness
and regularity theory is developed for such kinds of problems. Partial results assuming special structure of operators B and A can be found e.g. in
[4, 19]. However, these results are not applicable if B does not take the subgradient structure, which is the case of the coupled system (1.1)(1.2) due to
a non-symmetry in the parabolic term. Moreover, under rather general assumptions on b, the equation (1.1) degenerates in the parabolic part. Finally,
it is worth noting that further diculty lies in the convective term in the heat
equation, which represents strong nonlinearity in the model. Hence, we deal
with the nonlinear degenerate system under critical growth assumptions. In
the present paper we prove the existence of the weak solution to (1.1)(1.9)
and rigorously derive the corresponding homogenized problem letting 0
and using the two-scale convergence theory (in the sense of [40], see also [3]).
As far as we know, this is the rst attempt to carry out such an analysis for
(1.1)(1.9).
A brief bibliographical survey. Qualitative properties of particular problems
like (1.1), (1.3), (1.5) and (1.7), such as existence, uniqueness and regularity, have been studied by several authors, see e.g. [4, 19]. More recently,
in [12], the authors studied homogenization of the decoupled nonlinear degenerate parabolic problem (1.1), (1.3) and (1.7) with b(y, s) = s, however,
including more general nonlinear operator q = q (y, s, s). The operator
q was assumed to be periodic in y = x/ and degenerated in s. The
authors derived the limit equation letting 0 using the two-scale convergence theory and proven a strong convergence (corrector results). Similar
3

results can be found in [13], where the same authors considered the homogenization of multi-scale degenerate problem like (1.1), (1.3) and (1.7) with
b(x/, s) = s and q = [A(x/)u + g(x/, u )]. The authors derived
the homogenized equation and present results on the rst order corrector. In
[37], the authors studied the homogenization of nonlinear parabolic equations
like (1.1) with mixed boundary conditions under restrictive assumptions on b
and with non-degenerate and monotone elliptic part excluding the degenerate
van Genuchten model. A similar problem has been considered in [25].
Homogenization of a coupled system of diusion-convection equations in
a domain with periodic microstructure, modeling the ow of isothermal immiscible compressible uids through porous media, was theoretically studied
e.g. in [5, 6]. Homogenization of complex chemical processes in porous media has been recently studied e.g. in [18, 28, 29, 30]. Particularly, in [18],
the authors rigorously derived upscaled reaction-diusion models based on
a system of coupled semi-linear microstructural reaction-diusion equations
modeling concrete corrosion in sewer pipes. In [29], the authors applied two
scale convergence techniques to obtain the macro model of thermal-diusion
reaction problems (including the Dufour ad Soret eects) modeling e.g. behavior of concrete at high temperatures, drug delivery in biology tissues etc.
Just recently, in [28], the authors proved results on the weak solvability and
homogenization of a microscopic semi-linear elliptic systems modeling diffusion processes in porous media taking into account surface and chemical
reactions.
In the present paper, following the results in [13] concerning the isothermal degenerate problem, we intend to exploit the specic physical structure
of the transport model to obtain homogenized equations based on the twoscale convergence theory [3, 40]. To the best of our knowledge, mathematical
homogenization of degenerate coupled-transport problems in partially saturated porous media remains largely unexplored and remains open due to the
nonlinear coupling and degeneracies of the system.
Applications. The use of mathematical models in the simulation of heat and
moisture ow through heterogeneous porous media has received substantial
attention in recent years. Research eort has concentrated on the simulation of geothermal reservoirs, the eect of degradation due to the presence of
moisture in building and engineering structures, heat and moisture transport
in fresh concrete, heat resistance of concrete in re or under nuclear accidents
and many other common or extreme scenarios. Therefore advanced models
4

allowing to accurately analyze and forecast the coupled heat and mass transfer in porous media are of great importance. The main diculty lies in the
determination of several parameters used to dene relevant ecient transport
coecients of heterogeneous materials. In realistic situations, their experimental determination is rather complicated due to the complexity and lack
of enough reliable and accurate data. In virtue of the multi-scale character
of porous media, design of advanced engineering applications require the use
of materials with complex microstructures.
Several engineering models of coupled moisture ow and heat transport
in porous media possess a common structure, derived from two balance equations of mass of moisture and heat energy [7, 8, 35, 41, 42, 44] (neglecting
heat and moisture sources)
t ( w ) + (w v w ) = 0

(1.10)

t (cw w + cS S ) + (q h + cw w v w ) = 0.

(1.11)

and
In (1.10)(1.11), is a small scale parameter (the characteristic scale of heterogeneities) that express explicitly the multi-scale structure of the functions
, v w , q h , cS and S . Here, is the water content of the porous material,
is the temperature, v w represents the velocity of the moisture in the porous
material, q h stands for the conductive heat ux vector. Further, w is the
density of liquid water, cw represents the isobaric heat capacity of water, S
and cS , respectively, are the mass density and the isobaric heat capacity of
the solid. The amount of water present at a certain matric potential of the
porous medium is characterized by the water retention curve = (x, h)
(moisture sorption isotherms), where h is the so called matric potential.
Many natural and man-made porous media typically exhibit a hierarchical structure on many length scales. Moreover, the structural elements
themselves have structure. We assume that in each of these structures the
moisture ow is governed by Darcys law and the heat ow is described by
Fouriers law. Here the original moisture ow model is already a homogenized
model based on Stokes equations at the pore level (see e.g. [24, Chapter 3]
and the references therein). Our goal is to obtain the homogenized model,
valid at the macro level starting from the mesoscopic one. Hence, starting at
the mesoscale level, we apply the Darcys constitutive law for the mass ux
v w = k (x, , h) (h + g z ) ,
5

(1.12)

where g z stands for the vertical unit vector and k represents the hydraulic
conductivity of porous medium. Similarly, we assume the conductive heat
ux q h to be given by Fouriers law
q h = (x, , h)

(1.13)

with the thermal conductivity function . Combining Darcys law (1.12)

and the conservation of mass (1.10) yields the following equation
t (x, h) = (k (x, , h)h + g z k (x, , h)) .

(1.14)

Similarly, combining Darcys law (1.12), Fouriers law (1.13) and the conservation of energy (1.11) yields
t (cw w (x, h) + cS (x)S (x)) = ( (x, , h))
+ [cw w (k (x, , h)h + g z k (x, , h))] . (1.15)
The primary unknowns in the model are the absolute temperature = (x, t)
and the matric potential (the capillary pressure head) h = h(x, t) (singlevalued functions of the time t and the spatial position x).
Typically, neglecting residual water content, can be expressed as
(x, h) = s (x)SE (h),
where s is the saturated water content of the porous material and SE is
the eective saturation. Many parametric models have been proposed. The
models of Brooks and Corey [11] and of van Genuchten [21] are the most
popular for applications to air-water systems. The Brooks-Corey relationship
can be written as
SE (h) = |h| , h < 0,
where and are tting parameters related to the inverse of the entry
pressure head and the pore-size distribution. The model proposed by van
Genuchten [21] is given by the relationship
SE (h) = [1 + |h|n ]m ,

h < 0,

where , m and n are parameters. Retention curves in the saturated zone

can be assumed as [8]
SE (h) = sE h + 1,
6

h 0,

where sE > 0 is the specic (elasticity) storativity coecient.

The moisture and temperature dependence of hydraulic conductivity is
given by [16, 17, 44, 47]
k (x, , h) =

ks (x)kr (h)
,
w ()

where ks is the saturated hydraulic conductivity, kr stands for the h-dependent

relative hydraulic conductivity, which is commonly based on the power law
form [16, 44]

SE (h)Aw for h < 0,
kr (h) =
1
for h 0.
Here the exponent Aw is obtained from tting the data. Finally, w represents
the temperature dependent kinematic viscosity [51].
The eective thermal conductivity of partially saturated porous media
depends upon the degree of saturation with water and temperature. We
consider that the dependence may be approximated by a relationship
(x, , h) = d0 (x)d ()s (SE (h)),
where d0 , d and s are given functions obtained from tting the data.
Let us stress that, due to the structural complexity of highly porous
media, the coecient functions s , ks and d0 are assumed to be rapidly oscillating with large contrast.
 In our simplied heuristic approach we consider
a periodic composite Y = i Yi at the mesoscale level, in which the moisture
ow equation (1.14) as well as the heat equation (1.15) are valid in each component. Then the coecient functions s , ks and d0 become fast
oscillating
,
in Y and 
jump on the interfaces
of Yi having the form s (y) = i i (y)
si
, (y) =
. Here, i is the characteristic

(y)
k

(y)

ks (y) =
i
i
si
d0
d0i
i
i
are the individual characteristics of each
, k and
function of Yi and
si
si
d0i
component. Our goal is to obtain the upscaled equations, valid at the macro
level based on (1.14)(1.15), starting from the mesoscopic geometry described
above. The present asymptotic analysis is of topical interest in mathematical analysis [12, 13, 14, 23, 25, 36, 43], numerical homogenization, as well as
geotechnical and civil engineering (see e.g. [1, 34, 39, 45, 48, 49, 50]).
Kirchho transformation. It is worth noting that the nonlinear parabolic
equation (1.14) is degenerate in both elliptic and parabolic parts. A powerful
tool for handling the degenerate mass balance equation (1.14) is provided by
7

the so called Kirchho transformation, which employs the primitive function

: R R, = (), dened by

() =

kr (s)ds
0

and transforms degeneracies only to the parabolic term (see e.g. [4, 27]).
This formally leads to the system (1.1)(1.2), where
b(y, u) = s (y)SE ( 1 (u)),
1
a() =
,
w ()
c (y)S (y)
,
(y) = S
c w w
d ()s (SE ( 1 (u)))
(, u) =
,
cw w
g k (y, , 1 (u))
.
g(y, , u) = z
c w w
General assumptions on coecient functions b, a, and g are specied in
Section 2.
Remark 1.2. Let us note that, assuming a third type boundary conditions
(Robin conditions) for the temperature and the capillary pressure head, the
application of the Kirchho transformation leads to the more complicated
problem due to the coupled nonlinear boundary integral in the variational
formulation. Our existence and regularity result for system (1.1)(1.8) can
not be extended to such kinds of problems straightforwardly. More rigorous
qualitative study is required in this way.
Plan of the paper. In Section 2, we introduce necessary notation and definition of some function spaces, present some auxiliary results and specify
our assumptions on data and structure conditions on coecient functions
under which the main result of the paper is proven. In Section 3, we provide
the weak formulation of the initial-boundary value problem for the system
(1.1)(1.9). It is worth noting that, to make the text more readable, technical details of the existence proof for the multi-scale problem (1.1)(1.9) are
8

collected in Appendix A following the ideas coming from [10]. Theorem 3.5
represents the main result of the paper, derivation of the homogenized model
based on upscaling the problem (1.1)(1.9). The proof of the main result is
based on several preliminary results presented in Sections 4 and 5. In particular, in Section 4 we prove some a-priori estimates with respect to the scale
parameter , which are sucient for passing to the limit 0 in (1.1)(1.9)
using the two-scale convergence theory [40, 3], see Section 5. The proof of
the main result is completed in Section 6.
2. Preliminaries
Notation and denition of some function spaces. Throughout the paper, we
will always use positive constants C, c, c1 , c2 , . . . , which are not specied
and which may dier from line to line. I d denotes the identity matrix of size
2 2. We suppose p, q, p [1, ], p denotes the conjugate exponent to p,
p > 1, 1/p+1/p = 1. Lp () denotes the usual Lebesgue space equipped with
the norm

Lp () and W k,p (), k 0 (k need not to be an integer, see [31]),
denotes the usual Sobolev-Slobodecki space with the
. Recall
 norm

W k,p ()


that W 0,p () = Lp (). Further, we dene V := W 1,2 ();  = 0 .
D

We denote by V the dual space of V , (, ) stands for the inner product on

L2 () and ,  represents the duality pairing between V and V . Let X be
a Banach space. By Lp (0, T ; X) we denote the usual Bochner space (see [2]).
In the paper we shall use the following embedding theorems (see [2, 31]):
k,p
W ()  Lq (),

Lq () c

W k,p () , 1 q < +, kp = 2,

k,p
W ()  Lq (),

Lq () c

W k,p () , 1 q 2 p/(2 kp),

kp < 2,

k,p
W ()  L (),

L () c

W k,p () , kp > 2
(2.1)
1,2
(Y) be the space of elements of W 1,2 (Y)
for every W k,p (). Let Wper
having the same trace on opposite face of Y. Lpper (Y) is the subspace of Lp (Y)
of Y-periodic functions , i.e. (x + kei ) = (x) a.e. on R2 , for all k Z
and i {1, 2}, where {e1 , e2 } is the canonical basis of R2 . By Cper (Y) we
denote the Banach space of continuous and Y-periodic functions. Lp (; X)
stands for the set of measurable functions : x (x) X such
that
(x)
X Lp (). L2per (Y; C()) represents the space of measurable
functions : y Y (y) C() such that
(y)
C()) L2per (Y).
9

Structure and data properties. We begin by introducing our assumptions on

coecient functions in (1.1)(1.9).
(A1) b(y, ) is a positive and continuous function, Y-periodic with respect to
the rst variable (with a basic unit cell (0, 1)2 ) and strictly monotone
and Lipschitz continuous in such that
y b(y, ) C(R2 R)2 ,
0 < b(y, ) Cb , 0 < b(y, ) Cb (1 + ||)
(b(y, 1 ) b(y, 2 )) (1 2 ) > 0

(Cb = const),
(2.2)

for all y R2 and , 1 , 2 R, 1 = 2 . Moreover, there exists y R2

such that
(b(y, 1 ) b(y, 2 )) (1 2 ) (b(
y , 1 ) b(
y , 2 )) (1 2 )

(2.3)

for all y R2 and 1 , 2 R.

(A2) Functions and k , k = 1, 2, are smooth Y-periodic functions on R2 ,
such that (y) 0 > 0 and k (y) 0 > 0 for all y R2 , 0 , 0 =
const.
(A3) a and are continuous functions satisfying
0 < a1 a() a2 < +
0 < 1 (, ) 2 < +

(a1 , a2 = const),
(1 , 2 = const)

(2.4)

for all , R.
g : R2 R R R2 is a smooth vector function, Y-periodic with
respect to the rst variable with the basic unit cell (0, 1)2 such that
|g(y, , )| c < + (c = const)
for all y R2 and , R.
(A4) (Initial data) We assume u0 , 0 V L ().

10

Auxiliary results.
Remark 2.1 ([4], Section 1.1). Let us note that (A1) implies that there
is a C 1 -function : R2 R R, (y, 0) = 0, (y, )|=0 = 0, (strictly)
convex in R such that b(, ) b(, 0) = (, ) for all R. Introduce
the Legendre transform
 1

B(y, ) =
(b(y, ) b(y, s))ds =
(b(y, ) b(y, s))ds, y R2 .
0

Let us present some properties of B (see [4] for more details):

 1
(b(y, z) b(y, sz))z ds 0
z R, y R2 ,
B(y, z) =
0

B(y, z) B(y, r) (b(y, z) b(y, r))r

r, z R, y R2 ,
[b(y, z) b(y, 0)]z (y, z) = B(y, z) [b(y, z) b(y, 0)]z z R, y R2 .
Further, note that (A1) implies
B(y, z) B(
y , z)

(2.5)

for all y R2 and z R, where y is taken from (2.3).

Proposition 2.2 (see [4], Lemma 1.5.). Suppose (A4). Let > 0 and
u L2 (0, T ; V ), such that
b(x/, u) L (0, T ; L1 ()),

t b(x/, u) L2 (0, T ; V )

and


T
0


t b(x/, u), dt +

(b(x/, u) b(x/, u0 ))t dxdt = 0

for every test function L2 (0, T ; V ) W 1,1 (0, T ; L ()) with (T ) = 0.

Then B(x/, u) L (0, T ; L1 ()) and for almost all t the following formula
holds


 t
B(x/, u(t))dx B(x/, u0 )dx =
s b(x/, u), uds.
(2.6)

11

3. Main result
We rst formulate the problem (1.1)(1.8) in a weak sense.
Denition 3.1. Let > 0 be given. We say that a pair [u , ], such that
u L2 (0, T ; V )

and

L2 (0, T ; V ) L (T ),

is a weak solution of the system (1.1)(1.8) i

(i) b (, u ) L (0, T ; L1 ()), t b (, u ) L2 (0, T ; V ), t [(b (, u ) + ) ]
L2 (0, T ; V ) and


T
0


t b (x/, u ) , dt +

[b (x/, u ) b (x/, u0 )] t dxdt = 0

for every test function L2 (0, T ; V )W 1,1 (0, T ; L ()) with (T ) =

0 and
 T
t [(b (x/, u ) + (x/)) ] , dt
0

[(b (x/, u ) + (x/)) (b (x/, u0 ) + (x/)) 0 ] t dxdt
+
T

=0
for every test function L2 (0, T ; V )W 1,1 (0, T ; L ()) with (T ) =
0.
(ii) the couple [u , ] satises the following system


t b (x/, u ) , dt
0
+
[1 (x/)a( )u + g(x/, , u )] dxdt
T

=0

(3.1)

12

and


t [(b (x/, u ) + (x/)) ] ,  dt

0
+
2 (x/)( , u ) dxdt
T
[1 (x/)a( )u + g(x/, , u )] dxdt
+
T

=0

(3.2)

for all test functions L2 (0, T ; V ) and L2 (0, T ; V ).

Theorem 3.2. Let the assumptions (A1)(A4) be satised. Then there exists at least one weak solution of the system (1.1)(1.8). Moreover,
ess sup | | C,
T

(3.3)

where the constant C is independent of .

Proof. The existence of the weak solution to the problem like (1.1)(1.8)
with b = b(u), g = g(, u), 1 2 = 1 and = const is studied e.g. in
[10, 41]. The proof of Theorem 3.2 follows similar arguments, see Appendix
A for more details. 
Before we state the main result of our paper, we introduce the following
denition of the two-scale convergence, see [3].
Denition 3.3. Let 1 < q < +. A sequence v Lq () is said to twoscale converge to a function v Lq ( Y) if ( 0)

 

v (x, x/)dx
v(x, y)dydx
(3.4)

for all Lq (; Cper (Y)), q  = q/(q 1).

Remark 3.4. [15, Remark 9.4] Due to density properties, it is easily seen
that if v two-scale converges to v, convergence (3.4) (with q = 2) holds
also for any L2per (Y; C()) as well as for any of the form (x, y) =
1 (y)2 (x, y) with 1 L (Y) and 2 L2per (Y; C()).
13

The main result of this paper is summarized in the following theorem.

Theorem 3.5 (Main result). Let > 0. Suppose the data satisfy (A1)
(A4) and [u , ] is the couple according to Denition 3.1, i.e. the weak
solution of (1.1)(1.8). Then there exist the pairs [u, u1 ] and [, 1 ],
u L2 (0, T ; V ),

1,2
u1 L2 (T ; Wper
(Y)),

L2 (0, T ; V ) L (T ),

1,2
1 L2 (T ; Wper
(Y)),

and a sequence [uj , j ] such that limj+ j = 0+ and

uj  u
u j u
uj x u + y u1

weakly in L2 (0, T ; W 1,2 ()),

almost everywhere in T ,
in the two-scale sense

j
j
j
j

weakly in L2 (0, T ; W 1,2 ()),

weakly star in L (T ),
almost everywhere in T ,
in the two-scale sense.

and



x + y 1

Further, u and satisfy (in a weak sense) the following coupled homogenized
problem
t b (u) + q = 0
t (b (u) + ) = ( (, u, x, t) q )
u=0
=0

q n=0
n = 0
u(x, 0) = u0 (x)
(x, 0) = 0 (x)
where

q = (A (, x, t)u + g (, u, x, t)) .

14

in T ,
in T ,
in DT ,
in DT ,
in N T ,
in N T ,
in ,
in ,

(3.5)
(3.6)
(3.7)
(3.8)
(3.9)
(3.10)
(3.11)
(3.12)

(3.13)

In (3.5), (3.6) and (3.13), the homogenized coecient functions are

as


b (s) =
b(y, s) dy,
Y


(y) dy,
=
Y


A (r, x, t) = a(r) 1 (y) (I d + y w(x, y, t)) dy,

Y


(a(r)1 (y)y (x, y, t) + g(y, r, s)) dy,

g (r, s, x, t) =
Y


2 (y)(r, s) (I d + y (x, y, t)) dy,

(r, s, x, t) =

dened

(3.14)
(3.15)
(3.16)
(3.17)
(3.18)

where w = (w1 , w2 ), y w is the matrix (y w)ij = wi /yj , i, j = 1, 2.

Similarly, = (1 , 2 ) and y is the matrix (y )ij = i /yj , i, j =
1,2
1, 2. Further, wi , and i Wper
(Y) are periodic solutions of the following
cell problems:
y (1 (y)(ei + y wi )) = 0

wi dy = 0,

in Y, i = 1, 2,

(3.19)
(3.20)

further
y (a(r)1 (y)y ) = y g(y, r, s)

dy = 0,

in Y,

(3.21)
(3.22)

and, nally,
y (2 (y)(ei + y i )) = 0

i dy = 0.

in Y, i = 1, 2,

(3.23)
(3.24)

Schedule of the proof. In the proof of the main result we use the singleand two-scale convergence theory. First we derive suitable a priori estimates
for the solution [u , ] of the two-scale problem (1.1)(1.8), see Section 4,
key estimates (4.2), (4.5), (4.6), (4.12), (4.13) and (4.17). Based on these
15

estimates together with (3.3) we can pass to the limit as the space period
vanishes, and prove the convergence to a solution of a purely coarse-scale
formulation, recall (3.5) and (3.6). The two-scale convergence is carefully
studied in Section 5 and the proof of the main result is completed in Section 6
proving (3.19)(3.24).
4. A-priori estimates.
Here we present some a-priori bounds for the solution of (3.1)(3.2). First
we derive some estimates for u . Setting (0,t) u as a test function in (3.1)
((0,t) is the characteristic function of (0, t)) and using the integration by
parts (2.6) we arrive at


B (x/, u (t)) dx +

1 (x/)a( )|u |2 dxds



B (x/, u0 ) dx
g(x/, , u ) u dxds
=
t

for almost all t (0, T ). Applying the Youngs inequality to the second
integral on the right hand side and using (A3) we get

 t
B (x/, u (t)) dx +

u (s)
2W 1,2 () ds C

for almost all t (0, T ). Hence




sup
B (x/, u (t)) dx +
0tT

T
0

u (t)
2W 1,2 () dt C

(4.1)

and thus

u
L2 (0,T ;W 1,2 ()) C.

(4.2)

Next we proceed analogously as in [19]. Using (t,t+h) w, w V , as a test

function in (3.1) we obtain


t+h

1 (x/)a( )u w dxds
b (x/, u (t + h)) b (x/, u (t)) , w +

t
 t+h 
=
g(x/, , u ) w dxds. (4.3)

16

Now we set w = u (t + h) u (t) and integrate (4.3) with respect to t over

(0, T h) to obtain


T h

(b (x/, u (t + h)) b (x/, u (t)) , u (t + h) u (t)) dt

 T

Ch

u
2W 1,2 () + 1 dt
0

and owing to (4.1) we have



T h
0

(b (x/, u (t + h)) b (x/, u (t)) , u (t + h) u (t)) dt Ch. (4.4)

Now using (A1) there exists y R2 such that



T h
0

(b (
y , u (t + h)) b (
y , u (t)) , u (t + h) u (t)) dt Ch.

Further, combining (4.1) and (2.5) we deduce


sup
B (
y , u (t)) dx C.
0tT

(4.5)

(4.6)

Note that C is independent of . From (4.2) it follows that

u  u

weakly in L2 (0, T ; W 1,2 ())

(4.7)

(through a subsequence which we shall again denote by {}). Using the

compactness arguments of [4, Lemma 1.9] (see also [19, Eqs. (2.10)(2.12)])
and the estimates (4.5) and (4.6) we get
b(
y , u ) b(
y , u) in L1 (T )

(4.8)

and almost everywhere on T . Since b is strictly monotone (recall (2.2)), it

follows from (4.8) that, see [26, Proposition 3.35],
u u

almost everywhere on T .

(4.9)

Now let us derive some uniform estimates for . First, let > 0 and
dene u (t) = u0 and (t) = 0 for < t < 0. Letting 0 in the
17

estimate
 
1 t

[b (x/, u (s)) b (x/, u (s ))] (s)2 dxds

0
 
2 t
+
[b (x/, u (s)) (s) b (x/, u (s )) (s )] (s) dxds
0
 
2 t
+
[(x/) (s) (x/) (s )] (s) dxds
0

1 t
=
b (x/, u (s)) b (x/, u (s )) , (s)2  ds
0

2 t
+
b (x/, u (s)) (s) b (x/, u (s )) (s ), (s) ds
0

2 t
+
(x/) (s) (x/) (s ), (s) ds
0

1 t 2

(s) [b (x/, u (s)) + (x/)] ds

t

1 0 2
(s) [b (x/, u (s)) + (x/)] ds

we obtain



(t) [b (x/, u (t)) + (x/)] dx 02 [b (x/, u0 ) + (x/)] dx

 t

s b(x/, u ), (s)2 ds
0
 t
+
s ([b (x/, u (s)) + (x/)] (s)) , 2 (s) ds
(4.10)
0

for almost all t (0, T ). Now, using = [ ]2 as a test function in (3.1)

and = 2 in (3.2) and combining both equations together with (4.10) we
arrive at



(t) [b (x/, u (t)) + (x/)] dx +

22 (x/)( , u )| |2 dxds

t

02 (b (x/, u0 ) + (x/)) dx

18

for almost all t (0, T ). Hence, using (A3) we arrive at the inequality


(t)2 (b (x/, u (t)) + (x/)) dx



+ 21
| |2 dxds
02 (b (x/, u0 ) + (x/)) dx.
t

Therefore we conclude that

sup

0tT

(t)
2L2 ()

+
0

2W 1,2 () dt C,

(4.11)

which immediately yields

L2 (0,T ;W 1,2 ()) C.

(4.12)

It follows that converges weakly in L2 (0, T ; W 1,2 ()) to (along a selected

subsequence). To get almost everywhere convergence of we prove the
following estimate
 T h 
| (t + h) (t)|2 dxdt Ch,
(4.13)
0

which can be obtained similarly as (4.4). Namely, using (t,t+h) w, w V , as

a test function in (3.2) we obtain
[b (x/, u (t + h)) + (x/)] (t + h) [b (x/, u (t)) + (x/)] (t), w
 t+h 
2 (x/)( , u ) w dxds
+

t
 t+h 
+
[a(x/, )u + g(x/, , u )] w dxds
t

= 0.

(4.14)

Now we set w = (t + h) (t) and integrate (4.14) with respect to t over

19

(0, T h) to obtain


T h
0

(b (x/, u (t + h)) [ (t + h) (t)] , (t + h) (t)) dt

T h

([b (x/, u (t + h)) b (x/, u (t))] (t), (t + h) (t)) dt

 T h 
+
(x/)| (t + h) (t)|2 dxdt

0
 T h  t+h 
2 (x/)( (s), u (s)) (s) ( (t + h) (t)) dxdsdt
+

0
t
 T h  t+h 
+
(s) [a(x/, (s))u (s) + g(x/, (s), u (s))]
+

( (t + h) (t)) dxdsdt

= 0.
Hence

c1


T h

0
T h

| (t + h) (t)|2 dxdt

([b (x/, u (t + h)) b (x/, u (t))] (t), (t + h) (t)) dt

 T


2W 1,2 () +

L () (
u
W 1,2 () + 1)

W 1,2 () dt.
+ Ch
0

(4.15)

The rst integral on the right-hand side in (4.15) can be further estimated
using (A1) (Lipschitz continuity of b in the second variable), (3.3) and the

20

Youngs inequality to get

 T h
([b (x/, u (t + h)) b (x/, u (t))] (t), (t + h) (t)) dt
0
 T h 
2
C()

L (T )
|b (x/, u (t + h)) b (x/, u (t)) |2 dxdt

0
 T h 
| (t + h) (t)|2 dxdt
+
0

 T h 
C()
(b (x/, u (t + h)) b (x/, u (t)) , u (t + h) u (t)) dxdt

0
 T h 
| (t + h) (t)|2 dxdt.
(4.16)
+
0

Now, choosing suciently small, combining (4.15), (4.16) and using (3.3),
(4.2), (4.4) and (4.12) we get (4.13).
Finally, from (3.2), using (A2), (A3), (4.2), (4.12) and (3.3), we can write

t (b (x/, u ) + (x/) ))
L2 (0,T ;V ) C.

(4.17)

As a consequence of the preceding a priori estimates, in particular, (3.3),

(4.2), (4.5), (4.6), (4.11), (4.12), (4.13) and (4.17), we see that there exist
functions u, L2 (0, T ; W 1,2 ()), , L2 (T ) and the functionals ,
L2 (0, T ; V ), such that, along a selected subsequence denoted again by {},
we have
u  u
u u
b (x/, u ) 
t b (x/, u ) 

weakly in L2 (0, T ; W 1,2 ()),

almost everywhere in T ,
weakly in L2 (T ),
weakly in L2 (0, T ; V )

(4.18)
(4.19)
(4.20)
(4.21)

and




b (x/, u ) + (x/) 
t [b (x/, u ) + (x/) ] 

weakly in L2 (0, T ; W 1,2 ()),

almost everywhere in T ,
weakly in Lp (T ), 1 < p < +,
weakly star in L (T ),
weakly in L2 (T ),
weakly in L2 (0, T ; V ).
21

(4.22)
(4.23)
(4.24)
(4.25)
(4.26)
(4.27)

5. Two-scale limit
First we review basic properties of the two-scale (in the sense of [3, 40])
and single-scale convergence, see e.g. [24, Appendix A] and [15, Chapter 9]. We then state some results concerning the weak limits of the nonlinear
parabolic parts in (1.1) and (1.2) which will be used in Subsection 5.2. The
main result of this section is Theorem 5.5.
5.1. Auxiliary results
Theorem 5.1. [3, Chapter 2] Let v be a bounded sequence in W 1,p () with
1 < p < + that converges weakly to a limit v in W 1,p (). Then, v two1,p
scale converges to v(x), and there exists a function v1 Lp (; Wper
(Y)) such
that, up to a subsequence,
v x v + y v1

in the two-scale sense.

Lemma 5.2. [15, Lemma 9.1] Let Lp (; Cper (Y)) with 1 p < +.
Then the function (/, ) Lp () with

(/, )
Lp ()
(, )
Lp (;Cper (Y))
and (/, ) converges weakly in Lp () to

(y, x) dy.
Y

Lemma 5.3. [33, Lemma 4.2] Let (M, ) be a -nite measure space and
fn and gn L1 (M) be two sequences of functions, and let f, g, h L1 (M)
such that (as n +)
fn f
gn  g
fn gn  h

almost everywhere in M,
weakly in L1 (M),
weakly in L1 (M).

Then h = f g almost everywhere in M.

Lemma 5.4. There exists a subsequence of {}, still denoted by {}, such
that
b (x/, u )  b (u)
t b (x/, u )  t b (u)
b (x/, u ) + (x/)  b (u) +
22

weakly in L2 (T ),
weakly in L2 (0, T ; V ),
weakly in L2 (T )

(5.1)
(5.2)
(5.3)

and
t [b (x/, u ) + (x/) ]  t [b (u) + ]
weakly in L2 (0, T ; V ). (5.4)
Proof. Limits (5.1) and (5.2) are proven in [37, Corollary 4.3]. The limit
(5.3) follows from Lemma 5.3, (5.1) and the estimate




c,
b (x/, u ) + (x/) 
L2 (T )

where c is independent of . Now, (5.4) follows from (5.3) and (4.27). 

5.2. Passage to the limit for 0
1,2
(Y)) and a subseTheorem 5.5. There exist a function u1 L2 (T ; Wper

quence u (still denoted by u ) such that

u x u + y u1 (x, y, t)

in the two-scale sense.

1,2
(Y)) and a subsequence
Similarly, there exist a function 1 L2 (T ; Wper

(still denoted by ) such that

x + y 1 (x, y, t)

in the two-scale sense.

Further, the pairs (u, u1 ) and (, 1 ) satisfy the following two-scale homogenized coupled problem
 T 
 t b (y, u) dy, dt
0
 Y
1 (y)a() (x u + y u1 (x, y, t)) (x + y 1 (x, y, t)) dydxdt
+
T Y
 
+
g(y, , u) (x + y 1 (x, y, t)) dydxdt
T

=0

(5.5)

23

and
 T 
 t [b (y, u) + (y)] dy, dt
0
 Y
2 (y)(, u) (x + y 1 (x, y, t)) (x + y 1 (x, y, t)) dydxdt
+
T Y
 
1 (y)a() (x u + y u1 (x, y, t)) (x + y 1 (x, y, t)) dydxdt
+
T Y
 
+
g(y, , u) (x + y 1 (x, y, t)) dydxdt
T

=0

(5.6)

for all test functions , C0 (T ) and 1 , 1 C0 (T ; Cper

(Y)).

(Y)). We take test

Proof. Let , C0 (T ) and 1 , 1 C0 (T ; Cper
functions as
= (x, t) + 1 (x, x/, t)
(5.7)

and
= (x, t) + 1 (x, x/, t),

(5.8)

respectively, in (3.1) and (3.2). Note that, by (5.2) and the strong convergence of to in Lp (T ) we deduce


T
0


t b (x/, u ) , dt

T
0

t b (u), dt.

(5.9)

Similarly, by (5.4) we have



t [(b (x/, u ) + (x/)) ] ,  dt

0
 T

t [(b (u) + ) ] , dt.

0

Further, using (2.4), (4.18) and (4.23) we deduce

weakly in L2 (T )2 .

a( )u  a()u

24

(5.10)

Indeed, 1 L (Y) and L2per (Y; C()) so that 1 (x/) can be

used as a test function in the two-scale convergence of a( )u . Hence



1 (x/)a( )u dxdt =
a( )u [1 (x/) ] dxdt
T
T
 

1 (y)a() (x u + y u1 (x, y, t)) (x + y 1 (x, y, t)) dydxdt.

T

(5.11)
We deduce similarly that

g(x/, , u ) dxdt
T
 
g(y, , u) (x + y 1 (x, y, t)) dydxdt.

(5.12)

In the same manner we conclude that


2 (x/)( , u ) dxdt
T
 
2 (y)(, u) (x + y 1 (x, y, t)) (x + y 1 (x, y, t)) dydxdt

(5.13)
and

1 (x/)a ( ) u dxdt
T
 

1 (y)a() (x u + y u1 (x, y, t)) (x + y 1 (x, y, t)) dydxdt

T

(5.14)
and nally

g(x/, , u ) dxdt
T
 
g(y, , u) (x + y 1 (x, y, t)) dydxdt.

(5.15)

Thus, choosing (5.7) and (5.8), respectively, as test functions in (3.1) and
(3.2) the above convergences (5.9)(5.10) and (5.11)(5.15) are sucient for
taking the limit j 0 as j + (along a selected subsequence) to get
(5.5)(5.6). This completes the proof of Theorem 5.5. 
25

6. Proof of the main result

In this section we complete the proof of Theorem 3.5, the main result of
our paper. In particular, we identify the homogenized problem corresponding
to (1.1)(1.9), which can be obtained in the following way. Setting = 0 in
(5.5) we arrive at
 
1 (y)a() (x u + y u1 (x, y, t)) y 1 (x, y, t) dydxdt
T Y
 
=
g(y, , u) y 1 (x, y, t) dydxdt (6.1)
T

for all 1 C0 (T ; Cper

(Y)). Here we determine u1 (up to a constant) as

u1 (x, y, t) = x u(x, t) w(x, y, t) + (x, y, t),

(6.2)

where w(x, y, t), w = (w1 , . . . , wd ), and (x, y, t) can be obtained in the

following way. Indeed, combining (6.1) and (6.2) we arrive at
 
1 (y)a() (x u + y [x u(x, t) w(x, y, t)
T

+(x, y, t)]) y 1 (x, y, t) dydxdt

 
=
g(y, , u) y 1 (x, y, t) dydxdt
T

and thus


a()
 

+
T

[1 (y)x u (I d + y w(x, y, t))] y 1 (x, y, t) dydxdt

[1 (y)a()y (x, y, t) + g(y, , u)] y 1 (x, y, t) dydxdt

=0

(6.3)

(Y)). Here and in what follows, y w is the matrix

for all 1 C0 (T ; Cper
(y w)ij = wi /yj . Integrating by parts in (6.3) we deduce


a() y [1 (y)x u (I d + y w(x, y, t))]1 (x, y, t) dydxdt

T
Y
 
y [1 (y)a()y (x, y, t) + g(y, , u)] 1 (x, y, t) dydxdt

= 0.
26

From this, it is easy to show that w and can be obtained as solutions of

periodic cell problems (3.19)(3.20) and (3.21)(3.22).
Now let 1 = 0 in (5.5). Then we have
 T 
 t b (y, u) dy, dt
0

 Y
a() 1 (y) [x u + y u1 (x, y, t)] x dydxdt
+
T
Y
 
+
g(y, , u) x dydxdt
T

=0
and taking into account (6.2) we can write

 T
t b (y, u) dy, dt
0


 Y
x u(x, t)a()
1 (y) (I d + y w(x, y, t)) dy
x dxdt
+
T
Y




(a()1 (y)y (x, y, t) + g(y, , u)) dy x dxdt
+
T

=0

(6.4)

for all C0 (T ). Note that (6.4) represents the weak form of the equation
(3.5).
Similar procedure applies for the energy equation (5.6). Setting = 0 in
(5.6) leads to
 
2 (y)(, u) (x + y 1 (x, y, t)) y 1 (x, y, t) dydxdt
T Y


a()1 (y) (x u + y u1 (x, y, t)) y 1 (x, y, t) dydxdt
+
T Y
g(y, , u) y 1 (x, y, t) dydxdt
+
T

=0

(6.5)

(Y)). Here, 1 can be computed (up to a constant)

for all 1 C0 (T ; Cper
through the relationship

1 (x, y, t) = x (x, t) (x, y, t),

27

= (1 , . . . , d ),

(6.6)

which yields, substituting (6.2) and (6.6) into (6.5),

 
2 (y)(, u)x [I d + y (x, y, t)] y 1 (x, y, t) dydxdt
T Y


+
a() 1 (y)x u [I d + y w(x, y, t)] y 1 (x, y, t) dy dxdt
T
Y


0


[a()1 (y)y (x, y, t) + g(y, , u)] y 1 (x, y, t) dy dxdt
+
T
Y


0

=0

(6.7)

(Y)). Here and subsequently, y is the matrix

for all 1 C0 (T ; Cper
(y )ij = i /yj . Taking into account (3.19)(3.22), the second and third
integral on the left hand side in (6.7) vanish. Hence, i are obtained as the
unique solutions of the periodic cell problems (3.23)(3.24).
Now let 1 = 0 in (5.6). This leads to

 t [b (y, u) + (y)] dy, dt

 Y
2 (y)(, u) (x + y 1 (x, y, t)) x dydxdt
+
T Y
 
a()1 (y) (x u + y u1 (x, y, t)) x dydxdt
+
T Y
 
g(y, , u) x dydxdt
+
0

=0

28

for all C0 (T ) and, using (6.2) and (6.6), we can write

 

 T

t b (y, u) dy + (y)dy , dt
Y
0

  Y
2 (y)(, u)x (I d + y (x, y, t)) dy x dxdt
+
T
Y




1 (y)a()x u (I d + y w(x, y, t)) dy x dxdt

+
T
Y




(a()1 (y)y (x, y, t) + g(y, , u)) dy x dxdt

+
T

= 0.

(6.8)

We conclude that (6.8) is the weak form of the equation (3.6). The proof of
Theorem 3.5 is complete.
Appendix A. The existence of the weak solution to (1.1)(1.8)
Step 1. Approximations. Let us x N N and set = T /N be a time step.
We approximate our evolution problem by a semi-implicit time discretization
0
scheme. Let u0N u0 V L () and N
0 V L () be given. Then
n
n
we dene, in each time step, [uN , N ] as a solution of the following recurrence
problem:
n
Find a pair [unN , N
] V 2 , n = 1, . . . , N , such that

b(x/, unN ) b(x/, un1
N )
dx




n1
n1
)unN + g(x/, N
, unN ) dx
1 (x/)a(N
+

=0

(A.1)

for all V ;


n1
n1
n
n N
b(x/, unN )N
b(x/, un1
N )N
dx + (x/) N
dx

n1
n
+ 2 (x/)(N
, un1
N )N dx



n1
n1
n
+ N
)unN + g(x/, N
, unN ) dx
1 (x/)a(N

=0

(A.2)
29

for all V .
n1
2
Step 2. Existence of the solution to (A.1)(A.2). Let [un1
N , N ] V

2
L () be given and be small enough. Combining the standard existence
theory for elliptic problems, see e.g. [46, Chapter 2], [38, Chapter 3], [22, 32],
together with the regularity results in [20], the problem (A.1)(A.2) possesses
n
the solution [unN , N
] V 2 L ()2 .

Step 2. A-priori estimates. Here we prove some uniform estimates (with respect to N ) for the time interpolants of the solution. We dene the piecewise
constant interpolants (n = 1, 2, . . . , N ) N (t) = nN for t ((n 1), n ]
and, in addition, we extend N for t 0 by N (t) = 0 for t (, 0].
For a function we often use the simplied notation = (t), (t) =
)
, t (t) = (t+)(t) . Then, following (A.1)
(t ), t (t) = (t)(t

and (A.2), the piecewise constant interpolants uN L (0, T ; W 1,2 ())

L (0, T ; L ()), and N L (0, T ; W 1,2 ()) L (0, T ; L ()) satisfy the
equations

t b(x/, uN (t)) dx



+
uN (t) + g(x/, N (t ), uN (t)) dx
1 (x/)a(N (t ))

=0

(A.3)

for all V and t (0, T ) and




t b(x/, uN (t))N (t) + (x/)N (t) dx

+ 2 (x/)(N (t ), uN (t ))N (t) dx



+ N (t) 1 (x/)a(N (t ))
uN (t) + g(x/, N (t ), uN (t)) dx

=0

(A.4)

for all V and t (0, T ).

We test (A.3) with = uN (t) and integrate over t from 0 to s. For the

30

parabolic term we can write

 s
t b(x/, uN (t))
uN (t) dxdt

0


1 s

B(x/, uN (t)) B(x/, u0 ) dxdt.

s

(A.5)

Now, using (A.5) and standard estimates for the elliptic part and applying
the Gronwalls lemma we deduce, cf. [4],
 T

B(x/, uN (t))dx +
|
uN (t)|2 dxdt c.
(A.6)
sup
0tT

Hence it follows that

weakly in L2 (0, T ; V )

uN  u

(A.7)

along a selected subsequence (letting N +). Let k N and use

(t) = tk uN (s)
for j t (j + k) with (j 1) s j and 1 j T k, as a test
function in (A.3). Integrating over s and using (A.6) we obtain
 T k
(b(x/, uN (s + k )) b(x/, uN (s))) (
uN (s + k ) uN (s))ds ck.
0

(A.8)
Using the compactness arguments as in [4, Lemma 1.9] and [19, Eqs. (2.10)
(2.12)] we conclude
b(x/, uN ) b(x/, u) in L1 (T )

(A.9)

(for a subsequence) and almost everywhere on T . Since b is strictly monotone in the second variable, it follows from (A.9) that [26, Proposition 3.35]
uN u

almost everywhere.

(A.10)

The above a priori estimates furthermore imply

 T
 T
N , dt =
t b(x/, uN (t))(t) dxdt

0
0
 T h 
=
[b(x/, uN (t)) b(x/, u0 )] t (t) dxdt
0

31

for all L2 (0, T ; V ) with (t) = 0 for (a.a.) t > T h. From this we have
N 

weakly in L2 (0, T ; V )

(A.11)

and = t b (x/, u).

Now we use = 2N as a test function in (A.4) to obtain

2 t b(x/, uN (t))N (t)2 dx

+ 2 2 (x/)(N (t ), uN (t ))N (t) N (t) dx



1 (x/)a(N (t ))
uN (t) + g(x/, N (t ), uN (t)) N (t)N (t) dx
+2

= 0.

(A.12)

We are allowed to use (t) = N (t)2 as a test function in (A.3) to obtain


t b(x/, uN (t))N (t)2 dx



+
uN (t) + g(x/, N (t ), uN (t)) N (t)2 dx
1 (x/)a(N (t ))

= 0.

(A.13)

Combining (A.12) and (A.13) we deduce




t N (t)2 (b(x/, uN (t)) + (x/)) dx

2
1 
+
N (t) N (t ) (b(x/, uN (t )) + (x/)) dx


+ 2 2 (x/)(N (t ), uN (t ))|N (t)|2 dx

= 0.

(A.14)

Integrating (A.14) with respect to time t we obtain the a-priori estimate


 T
2

|N (t)| dx +

N (t)
2W 1,2 () dt c.
(A.15)
sup
0tT

This allows us to conclude that there exists L2 (0, T ; V ) such that, letting
N + (along a selected subsequence),
N 

weakly in L2 (0, T ; V ).
32

(A.16)

Further, in much the same way as in (A.8), we arrive at



T k
0

|w(x/, uN (s + k ), N (s + k )) w(x/, uN (s), N (s))|2 ds ck,

where w(x/, uN , N ) = [b(x/, uN ) + (x/)] N . From this we conclude that

 T k
|N (s + k ) N (s)|2 ds ck.
(A.17)
0

Hence
N

almost everywhere.
(A.18)

Let  be an odd integer. Using = N as a test function in (A.3) and
+1
= [/( + 1)]N
in (A.4) and combining both equations we have

N (t)+1 [b (x/, uN (t)) + (x/)] dx

+
N (s)1 2 (x/)(N (s), uN (s))|N (s)|2 dxds
t
0+1 [b (x/, u0 ) + (x/)] dx.
=

(A.19)

Recall that  is the odd integer. Hence the second integral in (A.19) is
nonnegative. Moreover, from (A.19) it follows that

N
L (0,T ;L+1 ()) C,

(A.20)

where the constant C is independent of , and . Now, let  + in

(A.20), we get

N
L (T ) C,
(A.21)
where the constant C is independent of and . Hence, letting N +
(along a selected subsequence), we have
N 

weakly star in L (T ).

(A.22)

Step 3. Passing to the limit. The above established convergences (A.7),

(A.10) and (A.16), (A.18) and (A.22) are sucient for taking the limit
N + in (A.1) and (A.2) (along a selected subsequence) to get the
weak solution of the system (1.1)(1.8) in the sense of Denition 3.1 and
satisfying (3.3). This completes the proof of Theorem 3.2.
33

Acknowledgment. The rst author of this work has been supported by the
16-20008S. The second author of this work has been supported
project GACR
by the Croatian Science Foundation (scientic project 3955: Mathematical
modeling and numerical simulations of processes in thin or porous domains).
[1] A. Abdulle, Y. Bai, Reduced-order modelling numerical homogenization.
Phil. Trans. R. Soc. A 372 (2014) 20130388.
[2] A. Adams, J.F. Fournier, Sobolev spaces, Pure and Applied Mathematics 140, Academic Press, 2003.
[3] G. Allaire, Homogenization of two-scale convergence, SIAM J. Math.
Anal. 23 (1992) 14821518.
[4] H.W. Alt, S. Luckhaus, Quasilinear elliptic-parabolic dierential equations, Math. Z. 183 (1983) 311341.
[5] B. Amaziane, S. Antontsev, L. Pankratov, A. Piatnitski, Homogenization of immiscible compressible two-phase ow in porous media: Application to gas migration in a nuclear waste repository, Multiscale Model.
Sim. 8 (2010) 20232047.
[6] B. Amaziane, M. Jurak, A. Vrbaski, Homogenization results for a coupled system modelling immiscible compressible two-phase ow in porous
media by the concept of global pressure, Appl. Anal. 92 (2013) 1417
1433.
[7] Z.P. Bazant, M.F. Kaplan, Concrete at high temperatures: material
properties and mathematical models, Concrete design and construction
series, Longman, 1996.
[8] J. Bear, Dynamics of uids in porous media, Amer. Elsevier, New York,
1979.
[9] M. Benes, L. Krupicka, On doubly nonlinear elliptic-parabolic systems
arising in coupled transport phenomena in unsaturated porous media,
Nonlinear Dier. Equ. Appl. 22 (2015) 121141.
[10] M. Benes, L. Krupicka, Weak solutions of coupled dual porosity ows in
fractured rock mass and structured porous media, J. Math. Anal. Appl.
433 (2016) 543565.
34

[11] R.H. Brooks and A.T. Corey, Properties of porous media aecting uid
ow, J. of the Irrigation and Drainage Division Proc. ASCE, 92 (1966)
6188.
[12] H. Cao, X. Yue, Homogenization of a Nonlinear Degenerate Parabolic
Dierential Equation, Acta. Math. Sin. 29 (2013) 14291436.
[13] H. Cao, X. Yue, Homogenization of Richards equation of van
Genuchten-Mualem model, J. Math. Anal. Appl. 412 (2014) 391400.
[14] Z. Chen, W. Deng, H. Ye, Upscaling of a class of nonlinear parabolic
equations for the ow transport in heterogeneous porous media, Comm.
Math. Sci. 3 (2005) 493515.
[15] D. Cioranescu, P. Donato, An Introduction to Homogenization, Oxford
University Press, 1999.
[16] F. Couture, W. Jomaa, J.R. Puiggali, Relative permeability relations: A
key factor for a drying model, Transport Porous Med. 23 (1996) 303335.
[17] M.L. Deangelis, E.F. Wood, A detailed model to simulate heat and
moisture transport in a frozen soil, IAHS-Aish P. 248 (1998).
[18] T. Fatima, N. Arab, E.P. Zemskov, A. Muntean, Homogenization of
a reactiondiusion system modeling sulfate corrosion of concrete in
locally periodic perforated domains, J. Eng. Math. 69 (2010) 261276.
[19] J. Filo, J. Kacur, Local existence of general nonlinear parabolic systems,
Nonlinear Anal. 24 (1995) 15971618.
[20] T. Gallouet, A. Monier, On the regularity of solutions to elliptic equations, Rendiconti Mat. 19 (1999) 471488.
[21] M. van Genuchten, A closed form equation for predicting the hydraulic
conductivity of unsaturated soil, Soil Sci. Soc. Am. J. 44 (1980) 892898.
[22] D. Gilbarg, N.S. Trudinger, Elliptic Partial Dierential Equations of
Second Order, Springer-Verlag Berlin Heidelberg, 2001.
[23] P. Heuser, Homogenization of Quasilinear EllipticParabolic Equations with Respect to Measures, Ruprecht-Karls-Universitat, Heidelberg, 2012.
35

[24] U. Hornung, Homogenization and porous media, Springer-Verlag New

York, 1997.
[25] H. Jian, On the homogenization of degenerate parabolic equations, Acta
Math. Appl. Sin.-E 16 (2000) 100110.
[26] J. Kacur, On a solution of degenerate elliptic-parabolic systems in
Orlicz-Sobolev spaces. I, Math. Z. 203 (1990) 153171.
[27] J. Kacur, R. Van Keer, Solution of degenerate parabolic variational
inequalities with convection, ESAIM-Math. Model. Num. 37 (2003) 417
431.
[28] V.A. Khoa, A. Muntean, Asymptotic analysis of a semi-linear elliptic
system in perforated domains: Well-posedness and correctors for the
homogenization limit, J. Math. Anal. Appl. 439 (2016) 271295.
[29] O. Krehel, T. Aiki, A. Muntean, Homogenization of a thermo-diusion
system with Smoluchowski interactions, Netw. Heterog. Media 9 (2014)
739762.
[30] O. Krehel, A. Muntean, P. Knabner, Multiscale modeling of colloidal
dynamics in porous media including aggregation and deposition, Adv.
Water Resour. 86 (2015) 209216.
[31] A. Kufner, O. John, S. Fuck, Function spaces, Academia, 1997.
[32] O.A. Ladyzhenskaya, N.N. Uraltseva, Linear and Quasilinear Elliptic
Equations, Academic Press New York and London, 1968.
[33] B. Li, W. Sun, Global existence of weak solution for nonisothermal
multicomponent ow in porous textile media, SIAM J. Math. Anal. 42
(2010) 30763102.
[34] K.F. Liu, Y.H. Wu, Y.C. Hsu, Homogenization theory applied to unsaturated solid-liquid mixture, J. Mech. 28 (2012) 329335.
[35] C.E. Majorana, V. Salomoni, B.A. Schreer, Hygrothermal and mechanical model of concrete at high temperature, Mater. Struct. 31 (1998)
378386.

36

[36] A. Mikelic, C. Rosier, Modeling solute transport through unsaturated

porous media using homogenization I, Comput. Appl. Math. 23 (2004)
195211.
[37] A.K. Nandakumaran, M. Rajesh, Homogenization of a nonlinear degenerate parabolic dierential equation, Electron. J. Di. Eq. 49 (2001)
119.
[38] J. Necas, Introduction to the Theory of Nonlinear Elliptic Equations,
Teubner-Texte zur Mathematik, Leipzig, 1983.
[39] I. Neuweiler, O.A. Cirpka, Homogenization of Richards equation in permeability elds with dierent connectivities, Water Resour. Res. 41
(2005) W02009.
[40] G. Nguetseng, A general convergence result for a functional related to
the theory of homogenization. SIAM J. Math. Anal. 20 (1989) 608623.
[41] S. Ning, Mathematical problems on the uid-solute-heat ow through
porous media, Acta Math. Sin. 6 (1990) 145157.
[42] S. Ning, An elliptic-parabolic system arising from the uid-solute-heat
ow through saturated porous media, Acta Math. Sin. 6 (1990) 224237.
[43] M. Peszy
nska, R. Showalter, S.-Y. Yi, Homogenization of a pseudoparabolic system, Appl. Anal. 88 (2009) 12651282.
[44] G.F. Pinder, W.G. Gray, Essentials Of Multiphase Flow And Transport
In Porous Media, John Wiley & Sons, 2008.
[45] Ph. Renard, G. de Marsily, Calculating equivalent permeability: A review, Adv. Water Resour. 20 (1997) 253278.
[46] T. Roubcek, Nonlinear Partial Dierential Equations with Applications,
Birkhauser, 2005.
[47] B.R. Scanlon, P.C.D. Milly, Water and heat uxes in desert soils: 2.
Numerical simulations, Water Resour. Res. 30 (1994) 721733.
[48] R.F. Sviercoski, A.W. Warrick, C.L. Winter, Two-scale analytical homogenization of Richards equation for ows through block inclusions,
Water Resour. Res. 45 (2009) W05403.
37


[49] J. S
ykora, T. Krejc, J. Kruis, M. Sejnoha.
Computational homogenization of non-stationary transport processes in masonry structures, J.
Comput. Appl. Math. 236 (2012) 47454755.

[50] J. S
ykora, M. Sejnoha,
J. Sejnoha.
Homogenization of coupled heat and
moisture transport in masonry structures including interfaces, Appl.
Math. Comput. 219 (2013) 72757285.
[51] H.R. Thomas, M.R. Sansom, Fully coupled analysis of heat, moisture
and air transfer in unsaturated soil, J. Engng. Mech. 121 (1995) 392405.

38